Search NASA⌕ Search

SEARCH · Search NASA

Results for “Parallelization efficiency”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 1,495 records · Page 83

The design and implementation of a parallel unstructured Euler solver using software primitives

This paper is concerned with the implementation of a 3D unstructured-grid Euler-solver on massively parallel distributed-memory computer architectures. The goal is to minimize solution time by achieving high computational rates with a numerically efficient algorithm. An unstructured multigrid algorithm with an edge-based data-structure has been adopted, and a number of optimizations have been devised and implemented in order to accelerate the parallel computational rates. The implementation is carried out by creating a set of software tools, which ease the implementation of computational problems on parallel architecture machines by relieving the user of the low-level machine specific issues. The quantitative effect of the various optimizations are demonstrated, and we show that the combined effect of these optimizations leads to roughly a factor of three performance improvement. The overall solution efficiency is compared with that obtained on the CRAY-YMP vector supercomputer.

Das, R.↗

Applying Machine Learning to Understand Write Performance of Large-scale Parallel Filesystems

In high-performance computing (HPC), I/O performance prediction offers the potential to improve the efficiency of scientific computing. In particular, accurate prediction can make runtime estimates more precise, guide users toward optimal checkpoint strategies, and better inform facility provisioning and scheduling policies. HPC I/O performance is notoriously difficult to predict and model, however, in large part because of inherent variability and a lack of transparency in the behaviors of constituent storage system components. In this work we seek to advance the state of the art in HPC I/O performance prediction by (1) modeling the mean performance to address high variability, (2) deriving model features from write patterns, system architecture and system configurations, and (3) employing Lasso regression model to improve model accuracy. We demonstrate the efficacy of our approach by applying it to a crucial subset of common HPC I/O motifs, namely, file-per-process checkpoint write workloads. We conduct experiments on two distinct production HPC platforms—Titan at the Oak Ridge Leadership Computing Facility and Cetus at the Argonne Leadership Computing Facility—to train and evaluate our models. We find that we can attain ≤ 30% relative error for 92.79% and 99.64% of the samples in our test set on these platforms, respectively.

Xie, Bing↗

A Distributed-Memory Package for Dense Hierarchically Semi-Separable Matrix Computations Using Randomization

In this paper, we present a distributed-memory library for computations with dense structured matrices. A matrix is considered structured if its off-diagonal blocks can be approximated by a rank-deficient matrix with low numerical rank. Here, we use Hierarchically Semi-Separable (HSS) representations. Such matrices appear in many applications, for example, finite-element methods, boundary element methods, and so on. Exploiting this structure allows for fast solution of linear systems and/or fast computation of matrix-vector products, which are the two main building blocks of matrix computations. The compression algorithm that we use, that computes the HSS form of an input dense matrix, relies on randomized sampling with a novel adaptive sampling mechanism. We discuss the parallelization of this algorithm and also present the parallelization of structured matrix-vector product, structured factorization, and solution routines. The efficiency of the approach is demonstrated on large problems from different academic and industrial applications, on up to 8,000 cores. Finally, this work is part of a more global effort, the STRUctured Matrices PACKage (STRUMPACK) software package for computations with sparse and dense structured matrices. Hence, although useful on their own right, the routines also represent a step in the direction of a distributed-memory sparse solver.

98 NUCLEAR DISARMAMENT, SAFEGUARDS, AND PHYSICAL P↗

The design and implementation of a parallel unstructured Euler solver using software primitives

This paper is concerned with the implementation of a three-dimensional unstructured grid Euler-solver on massively parallel distributed-memory computer architectures. The goal is to minimize solution time by achieving high computational rates with a numerically efficient algorithm. An unstructured multigrid algorithm with an edge-based data structure has been adopted, and a number of optimizations have been devised and implemented in order to accelerate the parallel communication rates. The implementation is carried out by creating a set of software tools, which provide an interface between the parallelization issues and the sequential code, while providing a basis for future automatic run-time compilation support. Large practical unstructured grid problems are solved on the Intel iPSC/860 hypercube and Intel Touchstone Delta machine. The quantitative effect of the various optimizations are demonstrated, and we show that the combined effect of these optimizations leads to roughly a factor of three performance improvement. The overall solution efficiency is compared with that obtained on the CRAY-YMP vector supercomputer.

Das, R.↗

RAPID Manufacturing Institute Final Report

The Rapid Advancement of Process Intensification Deployment (RAPID) Manufacturing Institute, founded in 2017, is a public/private partnership between the U.S. Department of Energy and the American Institute of Chemical Engineers (AIChE). RAPID promotes the development, deployment and commercialization of Process Intensification (PI) and Modular Chemical Process Intensification (MCPI) technologies, enabling U.S. manufacturing to reduce energy consumption, improve process efficiencies and lower investment and operating costs. This mission was carried out through parallel work breakdown structure elements including the establishment of committees to guide the operations and technical direction of RAPID, the establishment of management practices and institute processes, education and workforce development (EWD), and six technical focus areas for the development of technologies to advance PI and MCPI. Throughout the initial six-year cooperative agreement, RAPID worked to meet performance metrics which focused on the operation and sustainment of the institute, education and workforce development and the development of PI and MCPI for the advancement of U.S. manufacturing. All these metrics were successfully met through a total of 43 projects which leveraged $\$$70M Federal with $\$$90M cost share. As a result of these efforts, 84 private and public organizations were brought together by RAPID as members to co-invest in R&D, commercialization and deployment of innovative technologies. In the research portfolio, 82% of the 38 projects achieved > 20% energy efficiency improvement. A RAPID Test Network was developed with 51 testbed facilities to enable access to resources, facilities, tools, and expertise. Eight EWD programs were also developed with over 13,000 impressions. RAPID’s efforts to research, develop, demonstrate, and deploy high-impact PI and modular process technology solutions have enabled reduced energy use, increased sustainability, and improved profitability for U.S. manufacturing.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗

Cross-scale modeling of storm surge, tide, and inundation in Mid-Atlantic Bight and New York City during Hurricane Sandy, 2012

Driven by high-resolution NAM (North America Mesoscale) and ECMWF (European Centre for Medium-Range Weather Forecasts) atmospheric fields, a 3-D unstructured grid SCHISM (Semi-implicit Cross-scale Hydroscience Integrated System Model) was applied to simulate total water level in the Mid-Atlantic Bight during Hurricane Sandy in 2012. The simulated storm surge, tide, significant wave height, and wave peak period results compared favorably with NOAA observations along US East Coast, Long Island Sound, and New York Harbor. The maximum total water level at The Battery in the New York Harbor was accurately simulated with an absolute error of less than 0.08 m (out of 2.9 m peak surge) and a timing difference less than 10 min, which includes the effects on the order of 0.1–0.3 m (5–10%) from the coastal setup induced by the surface waves. The scenario comparisons of (1) “NAM” versus “ECMWF”, (2) “2-D” versus “3-D”, and (3) “with” versus “without” wind wave model were examined. The 3-D barotropic model forced by ECMWF including the effects of wind waves performs the best, attributed to wave-induced radiation stress and reduction of bottom stress when the 3-D version is used. Simultaneously, the ELCIRC-Sub model, using a 5-m topography/bathymetry sub-grid and a regular 200-m resolution finite volume computational model grid, was developed to simulate the street-level inundation in New York City. The momentum and mass fluxes calculated by the coarser base grid model were effectively coupled with the sub-grid so that running a full-blown high-resolution base model is not required. The ELCIRC-Sub model uses MPI (Message Passing Interface) parallel computing to enlarge the coverage of the land surface and an efficient non-linear solver to improve the accuracy of the wetting-and-drying processes. The temporal comparisons of modeled water level with NOAA's tidal stations and USGS' rapid-deployed gauges showed overall performance with an average error of 0.1 m. Particularly, ELCIRC-Sub captured the profile of the highest peak surge (3.9 m), which rose 2 m within 3 hour at Kings Point, NY. The spatial comparisons of the modeled peak water level at 80 surveyed locations showed an average error less than 0.13 m. The modeled maximum inundation extent also matched well with an 80% hit rate against FEMA's Hurricane Sandy maximum flooding extent.

54 ENVIRONMENTAL SCIENCES↗

Long duration battery sizing, siting, and operation under wildfire risk using progressive hedging

Battery sizing and siting problems are computationally challenging due to the need to make long-term planning decisions that are cognizant of short-term operational decisions. This paper considers sizing, siting, and operating batteries in a power grid to maximize their benefits, including price arbitrage and load shed mitigation, during both normal operations and periods with high wildfire ignition risk. Here we formulate a multi-scenario optimization problem for long duration battery storage while considering the possibility of load shedding during Public Safety Power Shutoff (PSPS) events that de-energize lines to mitigate severe wildfire ignition risk. To enable a computationally scalable solution of this problem with many scenarios of wildfire risk and power injection variability, we develop a customized temporal decomposition method based on a progressive hedging framework. Extending traditional progressive hedging techniques, we consider coupling in both placement variables across all scenarios and state-of-charge variables at temporal boundaries. This enforces consistency across scenarios while enabling parallel computations despite both spatial and temporal coupling. The proposed decomposition facilitates efficient and scalable modeling of a full year of hourly operational decisions to inform the sizing and siting of batteries. With this decomposition, we model a year of hourly operational decisions to inform optimal battery placement for a 240-bus WECC model in under 70 min of wall-clock time.

25 ENERGY STORAGE↗

The finale of a trilogy: comparing terpolymers and ternary blends with structurally similar backbones for use in organic bulk heterojunction solar cells

Building on our previous works that compared the efficacy of terpolymers vs. ternary blends in improving the performance of bulk heterojunction organic solar cells, the final piece of this series of studies focuses on comparing terpolymer and ternary blends constructed with two polymers with structurally similar backbones (monoCNTAZ and FTAZ) yet markedly different open circuit voltage (V oc ) values. Furthermore, terpolymers and ternary blends of five different ratios were studied and the results demonstrate that while the overall performance of both the systems is similar, the ternary blends exhibit higher short circuit current (J sc ) values, while the terpolymers exhibit higher V oc values. Investigation of the charge transfer state using low-energy external quantum efficiency (EQE) indicates that the ternary blends are governed by a parallel-like mechanism, while the terpolymer does not follow this mechanism. Furthermore, the key morphological difference between the systems, as elucidated by resonance soft X-ray scattering (RSoXS), is the slightly smaller size (~60 nm) of domains in the ternary blends compared to that of the terpolymer (~80 nm), which may affect exciton harvesting in the terpolymer system and lead to lower J sc values. In addition, a lower driving force for the formation of charge transfer (CT) state is also likely to contribute to the lower J sc values in the terpolymer system. All together, the data show that structurally similar (perhaps even miscible) polymers still exhibit key differences in performance when paired in terpolymers vs. ternary blends and allow us to further illuminate the underlying mechanisms of such complex systems.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Breakdown assisted by a novel electron drift injection in the J-TEXT tokamak

A novel electron drift injection (EDI) system aiming to improve breakdown behavior has been designed and constructed on the Joint Texas EXperiment Tokamak Tokamak. Electrons emitted by the system undergo the E×B drift, ∇B drift and curvature drift in sequence in order to traverse the confining magnetic field. A local electrostatic well, generated by a concave-shaped plate biased more negative than the cathode, is introduced to interrupt the emitted electrons moving along the magnetic field line (in the parallel direction) in an attempt to bring an enhancement of the injection efficiency and depth. A series of experiments have demonstrated the feasibility of this method, and a penetration distance deeper than 9.5 cm is achieved. Notable breakdown improvements, including the reduction of breakdown delay and average loop voltage, are observed for discharges assisted by EDI. The lower limit of successfully ionized pressure is expanded.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗

Explicitly correlated second-order Møller-Plesset perturbation theory in a Divide-Expand-Consolidate (DEC) context

We present the DEC-RIMP2-F12 method where we have augmented the Divide Expand-Consolidate resolution-of-the-identity second-order Møller-Plesset perturbation theory method (DEC-RIMP2) [P. Baudin et al., J. Chem. Phys. 144, 054102 (2016)] with an explicitly correlated (F12) correction. Here, the new method is linear-scaling, massively parallel, and it corrects for the basis set incompleteness error in an efficient manner. In addition, we observe that the F12 contribution decreases the domain error of the DEC-RIMP2 correlation energy by roughly an order of magnitude. An important feature of the DEC scheme is the inherent error control defined by a single parameter, and this feature is also retained for the DEC-RIMP2-F12 method. In this paper we present the working equations for the DEC-RIMP2-F12 method and proof of concept numerical results for a set of test molecules.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

GPU-accelerated dislocation dynamics using subcycling time-integration

Discrete dislocation dynamics (DDD) simulations are being increasingly employed to investigate metal plasticity at the mesoscale. However, in spite of its ability to access beyond the length and time limits of atomistic methods, the DDD model is still limited by its high computational cost, with ranges of achievable strains too low and strain rates too high by several orders of magnitude compared with typical experimental conditions. By combining the efficiency of the recently developed subcycling time-integrator with the highly parallel architecture of graphics processing unit (GPU) devices, we developed a DDD model that provides significant acceleration compared to existing implementations. Our GPU-accelerated implementation enables large-scale DDD simulations that can reach relevant levels of strain using a moderate amount of computational resources.

36 MATERIALS SCIENCE↗

Electron transport in graphene/graphene side-contact junction by plane-wave multiple-scattering method

Electron transport in graphene is along the sheet but junction devices are often made by stacking different sheets together in a “side-contact” geometry which causes the current to flow perpendicular to the sheets within the device. Such geometry presents a challenge to first-principles transport methods. We solve this problem by implementing a plane-wave-based multiple-scattering theory for electron transport. In this study, this implementation improves the computational efficiency over the existing plane-wave transport code, scales better for parallelization over large number of nodes, and does not require the current direction to be along a lattice axis. As a first application, we calculate the tunneling current through a side-contact graphene junction formed by two separate graphene sheets with the edges overlapping each other. We find that transport properties of this junction depend strongly on the AA or AB stacking within the overlapping region as well as the vacuum gap between two graphene sheets. Finally, such transport behaviors are explained in terms of carbon orbital orientation, hybridization, and delocalization as the geometry is varied.

36 MATERIALS SCIENCE↗

Fault-Tolerant and Elastic Streaming MapReduce with Decentralized Coordination

The MapReduce programming model, due to its simplicity and scalability, has become an essential tool for processing large data volumes in distributed environments. Recent Stream Processing Systems (SPS) extend this model to provide low-latency analysis of high-velocity continuous data streams. However, integrating MapReduce with streaming poses challenges: first, the runtime variations in data characteristics such as data-rates and key-distribution cause resource overload, that inturn leads to fluctuations in the Quality of the Service (QoS); and second, the stateful reducers, whose state depends on the complete tuple history, necessitates efficient fault-recovery mechanisms to maintain the desired QoS in the presence of resource failures. We propose an integrated streaming MapReduce architecture leveraging the concept of consistent hashing to support runtime elasticity along with locality-aware data and state replication to provide efficient load-balancing with low-overhead fault-tolerance and parallel fault-recovery from multiple simultaneous failures. Our evaluation on a private cloud shows up to 2:8 improvement in peak throughput compared to Apache Storm SPS, and a low recovery latency of 700 -1500 ms from multiple failures.

97 MATHEMATICS AND COMPUTING↗

Architectures and Control of Submodule Integrated DC-DC Converters for Photovoltaic Applications

This paper describes photovoltaic (PV) module architectures with parallel-connected submodule-integrated dc-dc converters (subMICs) that improve efficiency of energy capture in the presence of partial shading or other mismatch conditions. The subMICs are bidirectional isolated dc-dc converters capable of injecting or subtracting currents to balance the module substring voltages. When no mismatches are present, the subMICs are simply shut down, resulting in zero insertion losses. It is shown that the objective of minimum subMIC power processing can be solved as a linear programming problem. A simple close-to-optimal distributed control approach is presented that allows autonomous subMIC control without the need for a central controller or any communication among the subMICs. Furthermore, the proposed control approach is well suited for an isolated-port architecture, which yields additional practical advantages including reduced subMIC power and voltage ratings. The architectures and the control approach are validated by simulations and experimental results using three bidirectional flyback subMICs attached to a standard 180-W, 72-cell PV module, yielding greater than 98% module-level power processing efficiency for a mismatch less than 25%.

Olalla, C↗

Accelerated fast iterative shrinkage thresholding algorithms for sparsity-regularized cone-beam CT image reconstruction

Purpose: The development of iterative image reconstruction algorithms for cone-beam computed tomography (CBCT) remains an active and important research area. Even with hardware acceleration, the overwhelming majority of the available 3D iterative algorithms that implement nonsmooth regularizers remain computationally burdensome and have not been translated for routine use in time-sensitive applications such as image-guided radiation therapy (IGRT). In this work, two variants of the fast iterative shrinkage thresholding algorithm (FISTA) are proposed and investigated for accelerated iterative image reconstruction in CBCT. Methods: Algorithm acceleration was achieved by replacing the original gradient-descent step in the FISTAs by a subproblem that is solved by use of the ordered subset simultaneous algebraic reconstruction technique (OS-SART). Due to the preconditioning matrix adopted in the OS-SART method, two new weighted proximal problems were introduced and corresponding fast gradient projection-type algorithms were developed for solving them. We also provided efficient numerical implementations of the proposed algorithms that exploit the massive data parallelism of multiple graphics processing units. Results: The improved rates of convergence of the proposed algorithms were quantified in computer-simulation studies and by use of clinical projection data corresponding to an IGRT study. The accelerated FISTAs were shown to possess dramatically improved convergence properties as compared to the standard FISTAs. For example, the number of iterations to achieve a specified reconstruction error could be reduced by an order of magnitude. Volumetric images reconstructed from clinical data were produced in under 4 min. Conclusions: The FISTA achieves a quadratic convergence rate and can therefore potentially reduce the number of iterations required to produce an image of a specified image quality as compared to first-order methods. We have proposed and investigated accelerated FISTAs for use with two nonsmooth penalty functions that will lead to further reductions in image reconstruction times while preserving image quality. Moreover, with the help of a mixed sparsity-regularization, better preservation of soft-tissue structures can be potentially obtained. The algorithms were systematically evaluated by use of computer-simulated and clinical data sets.

60 APPLIED LIFE SCIENCES↗

Self-action of Bessel wave packets in a system of coupled light guides and formation of light bullets

The self-action of two-dimensional and three-dimensional Bessel wave packets in a system of coupled light guides is considered using the discrete nonlinear Schrödinger equation. The features of the self-action of such wave fields are related to their initial strong spatial inhomogeneity. The numerical simulation shows that for the field amplitude exceeding a critical value, the development of an instability typical of a medium with the cubic nonlinearity is observed. Various regimes are studied: the self-channeling of a wave beam in one light guide at powers not strongly exceeding a critical value, the formation of the “kaleidoscopic” picture of a wave packet during the propagation of higher-power radiation along a stratified medium, the formation of light bullets during competition between self-focusing and modulation instabilities in the case of three-dimensional wave packets, etc. In the problem of laser pulse shortening, the situation is considered when the wave-field stratification in the transverse direction dominates. This process is accompanied by the self-compression of laser pulses in well enough separated light guides. The efficiency of conversion of the initial Bessel field distribution to two flying parallel light bullets is about 50%.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

3D_MolGNN_RL

3D-MolGNNRL, couples reinforcement learning (RL) to a deep generative model based on 3D-Scaffold to generate target candidates specific to a protein pocket building up atom by atom from the core scaffold. 3D-MolGNNRL provides an efficient way to optimize key features within a protein pocket using a parallel graph neural network model. The agent learns to build molecules in 3D space while optimizing the binding affinity, potency, and synthetic accessibility of the candidates generated for the SARS-CoV-2 Main protease

Kumar, Neeraj↗

Sage: parallel semi-asymmtric graph algorithms for NVRAMs

Non-volatile main memory (NVRAM) technologies provide an attractive set of features for large-scale graph analytics, including byte-addressability, low idle power, and improved memory-density. NVRAM systems today have an order of magnitude more NVRAM than traditional memory (DRAM). NVRAM systems could therefore potentially allow very large graph problems to be solved on a single machine, at a modest cost. However, a significant challenge in achieving high performance is in accounting for the fact that NVRAM writes can be much more expensive than NVRAM reads. In this paper, we propose an approach to parallel graph analytics using the Parallel Semi-Asymmetric Model (PSAM), in which the graph is stored as a read-only data structure (in NVRAM), and the amount of mutable memory is kept proportional to the number of vertices. Similar to the popular semi-external and semi-streaming models for graph analytics, the PSAM approach assumes that the vertices of the graph fit in a fast read-write memory (DRAM), but the edges do not. In NVRAM systems, our approach eliminates writes to the NVRAM, among other benefits. To experimentally study this new setting, we develop Sage, a parallel semi-asymmetric graph engine with which we implement provably-efficient (and often work-optimal) PSAM algorithms for over a dozen fundamental graph problems. We experimentally study Sage using a 48--core machine on the largest publicly-available real-world graph (the Hyperlink Web graph with over 3.5 billion vertices and 128 billion edges) equipped with Optane DC Persistent Memory, and show that Sage outperforms the fastest prior systems designed for NVRAM. Importantly, we also show that Sage nearly matches the fastest prior systems running solely in DRAM, by effectively hiding the costs of repeatedly accessing NVRAM versus DRAM.

97 MATHEMATICS AND COMPUTING↗